1 paper
Abhijit Kumar, Zoey Wu, Mohit Suley
Humans know when to reach for help e.g. 347×28 warrants a calculator while 2+2 does not. Language models do not. Prompt-based approaches can instruct a model when to inv…